github infiniflow/ragflow v1.0.0-rc2

latest release: nightly
5 hours ago

1.0.0-rc2 is the second release candidate for v1.0.0. This release fixes several issues in rc1 and introduces new features and expanded support.

Highlights

  1. Introduced Agentic mode for chat, replacing the Low, Medium, High, and Ultra thinking levels.
  2. Preserved compatibility with existing v1.0.0-rc1 data, allowing direct upgrades.
  3. Added support for Team/Me sharing.
  4. Added support for model providers DemonRoute and Opper, web search tools Firecrawl and Search1API, and the MinerU 4 model.
  5. Added retrieval support for Czech and Slovak.
  6. Unified the 12 built-in document parsers with the Pipeline parsers.
  7. Added Docker images for the arm64 architecture.

What's Changed

  • refactor(parser): unify spreadsheet tables on a segmented HTML wire by @jay77721 in #20358
  • refactor: use correct ragflow.conf as nginx config file by @JinHai-CN in #20321
  • docs:update Docker commands in deployment guides by @Yannnnnnny in #20353
  • feat(parser): add opt-in VLM enhancement for images by @jay77721 in #20166
  • refactor(parser_config): make all settings component-scoped (zero flat-key compat) by @xugangqiang in #20315
  • refactor: agent templates by @JinHai-CN in #20364
  • Fix: clear GetByID by @Lynn-Inf in #20371
  • refa: migrate provider model calls by @buua436 in #20355
  • fix(web): show api key before base url in every model provider form by @dcc123456 in #20361
  • Fix: Unable to delete the last item in the token chunker's delimiter configuration. by @cike8899 in #20376
  • refactor(parser): move native PDF/Office extraction into deepdoc facade by @xugangqiang in #20273
  • docs: align onnxruntime_go binding version in comments with go.mod (v1.29.0) by @xugangqiang in #20370
  • fix: route embeddings through model wrapper by @buua436 in #20378
  • Refactor: explictly use keywords_similarity_weight by @wangq8 in #20379
  • Update CI by @JinHai-CN in #20380
  • refactor: remove obsolete chat session compatibility API by @QAQadws in #20374
  • fix(mappings): support any embedding dimension on Elasticsearch by @qinling0210 in #20368
  • refactor: knowledge compilation templates by @JinHai-CN in #20383
  • refactor: remove unused py code by @JinHai-CN in #20388
  • refactor: remove unused sandbox code by @JinHai-CN in #20389
  • Refactor: vector_similarity_weight -> keywords_similarity_weight by @wangq8 in #20385
  • fix(document): clone parser_config to break dataset aliasing; drop dead parser-config helpers by @jay77721 in #20381
  • refactor(web): remove Python backend compatibility code by @dcc123456 in #20375
  • refactor: route audio models through wrappers by @buua436 in #20387
  • Fix Dataset ingestion overview by @QAQadws in #20190
  • refactor: remove unused code and doc by @JinHai-CN in #20392
  • Refactor: better message - canvas -> agent by @wangq8 in #20395
  • refactor(workflow): preserve completed parallel results by @6ba3i in #20391
  • fix(cli): drop PENDING as a parameter for message queue commands by @flyingwhit in #20367
  • refactor: move res into internal by @JinHai-CN in #20390
  • fix(file): keep polling until cleared datasets disappear from the list by @dcc123456 in #20423
  • Fix: better GetByID by @Lynn-Inf in #20421
  • fix(chunker): attach media context on General markdown and HTML paths (#20141) by @Souravrajvi0 in #20265
  • fix(parser): preserve DOCX XML character references by @jay77721 in #20359
  • Doc: fix errors by @JinHai-CN in #20427
  • chore(deps): bump hpack from 4.1.0 to 4.2.0 by @dependabot[bot] in #20434
  • Refactor: Delete rag by @wangq8 in #20425
  • chore(deps): bump fast-uri from 3.1.7 to 3.1.8 in /web by @dependabot[bot] in #20433
  • chore(deps): bump brace-expansion in /web by @dependabot[bot] in #20435
  • chore(deps): bump cryptography from 46.0.7 to 50.0.0 by @dependabot[bot] in #20437
  • refactor(workflow): bound parallel task admission by @6ba3i in #20399
  • fix(parser): wire MinerU server_url through Go PDF parser by @Souravrajvi0 in #19445
  • fix: Add the missing env configuration by @myf-bee in #20444
  • refactor: use model tokenizer for rerank by @buua436 in #20422
  • fix(parser): restore picture OCR independently of VLM enhancement by @jay77721 in #20424
  • Refactor: better to use component display name by @wangq8 in #20431
  • refactor: remove python implemention of whatsapp channel by @JinHai-CN in #20432
  • refa: route chat calls through ChatModel by @buua436 in #20441
  • Fix: Promote the visual enhancement option to a global setting. by @cike8899 in #20397
  • rename entrypoint-go.sh to entrypoint.sh by @myf-bee in #20448
  • refactor: remove unused python module by @JinHai-CN in #20436
  • refactor(deps): use Go dependency downloader as sole entrypoint by @jay77721 in #20360
  • refactor(canvas): reject failed parallel state clones by @6ba3i in #20443
  • refactor: remove unused go functions by @JinHai-CN in #20455
  • Drop unscoped (flat) parser_config keys instead of rejecting them by @xugangqiang in #20445
  • refactor(mcp): reject unvalidated redirect destinations by @6ba3i in #20447
  • refactor(agent): avoid unsafe HTTP retries by @6ba3i in #20456
  • refactor(agent): reject non-global destinations by @6ba3i in #20461
  • refactor(agent): bound local output capture by @6ba3i in #20464
  • refactor(components): remove the empty Answer stub by @6ba3i in #20454
  • chore(deps): bump sharp from 0.35.2 to 0.35.5 in /internal/channels/whatsapp-gateway by @dependabot[bot] in #20471
  • docs(web): remove stale UmiJS references after Vite migration by @sunshanhui in #20467
  • add nats monitoring port variable in docker-compose-base.yml by @myf-bee in #19664
  • Fix: Text with swapped keyword similarity weights by @cike8899 in #20472
  • Fix: The separator line causes the font size of the agent message error text to be too large. by @cike8899 in #20462
  • fix(metadata): show empty state when no metadata keys exist by @dcc123456 in #20428
  • refactor(chat): push pagination down to ListByOwnerIDs by @flyingwhit in #20450
  • refactor(canvas): preserve parallel item state across resumes by @6ba3i in #20440
  • fix: add json output alias on Retrieval node by @flyingwhit in #20394
  • Refactor: use type_kwd instead of knowledge_graph_kwd by @wangq8 in #20475
  • Port agentic RAG to Go, expose it as a chat mode, and add per-dialog failover by @yuzhichang in #20503
  • feat(agent,ingestion): carry the grep/bm25 contracts and index declared titles by @yuzhichang in #20542
  • chore: fix some function names in comments by @somasoil in #20550
  • chore(deps): bump urllib3 from 2.7.0 to 2.8.0 by @dependabot[bot] in #20538
  • docs: fix ui words by @Yannnnnnny in #20426
  • fix(citation): open external web documents directly by @dcc123456 in #20452
  • refactor(agent): preserve Stagehand client leases by @6ba3i in #20451
  • refa: simplify model resolution and rerank calls by @buua436 in #20457
  • docs: add DemonRoute to supported model providers by @Harsh28-raj in #20548
  • Refactor: remove API_PROXY_SCHEME by @Lynn-Inf in #20470
  • feat: add Opper as a model provider by @Felixkw12 in #20577
  • Doc: remove supported models by @JinHai-CN in #20584
  • fix(api): keep a parsing progress of 0.96 from being reported as 1.0 by @SashaMIT in #20414
  • docs: document the Firecrawl web search provider in the HTTP API reference by @JuampiHernandez in #20479
  • WorkerPool resize overspawns after shrinking an idle pool by @Taranum01 in #20490
  • fix(wiki): preserve external link navigation by @Mnilax in #20517
  • fix(go-agent): read Keenable result content from snippet by @errmakov in #20480
  • fix(elasticsearch): propagate search failures with index context by @quifox in #20546
  • docs: fix DOCX parser test link after move to deepdoc/parser/office by @pratikgx in #20521
  • fix(agent): remove obsolete download info toggle by @6ba3i in #20588
  • refactor(search): push pagination down to ListByOwnerIDs by @flyingwhit in #20386
  • fix(storage): URL-encode S3 and OSS copy sources by @AnnaSuSu in #20504
  • fix(cli): refresh filesystem providers when switching API servers by @helloxjade in #20498
  • fix(web): keep pending filter selections when the filter options refresh by @Xavierando in #20581
  • chore(deps): bump dompurify from 3.3.2 to 3.4.16 in /web by @dependabot[bot] in #20589
  • chore(deps): bump @vue/server-renderer and vue in /web by @dependabot[bot] in #20591
  • chore(deps): bump tinypool and oxfmt in /web by @dependabot[bot] in #20593
  • chore(deps): bump multidict from 6.7.0 to 6.9.1 by @dependabot[bot] in #20590
  • chore(deps): bump werkzeug from 3.1.8 to 3.1.9 by @dependabot[bot] in #20592
  • Fix: Clicking the batch parsing button for the dataset file produces no response. by @cike8899 in #20602
  • Refactor: knowledge compilation field name refactor by @wangq8 in #20587
  • fix(wiki): use plain-text summaries by @QAQadws in #20598
  • feat: add SiliconFlow model catalog entries by @buua436 in #20596
  • Go: go mod tidy by @JinHai-CN in #20605
  • fix: backfill tenant model max tokens in rc2 by @buua436 in #20601
  • Remove pullfrog by @JinHai-CN in #20615
  • fix(agent): port Sofya search tool to Go by @yusufgurdogan in #20410
  • fix: reject non-string metadata selector document IDs by @helloxjade in #20540
  • serenedb: stop compiling the similarity threshold into the ANN scan by @deadtrickster in #20222
  • perf(tokenizer): memoize term lookups and precompile per-token regexes by @qinling0210 in #20610
  • i18n(web): clarify agent citation switch label across locales by @dcc123456 in #20607
  • Fix: Align the content of the th element within the chunk to the left. by @cike8899 in #20609
  • Fix: disable agent citation, should not show citation files by @wangq8 in #20616
  • fix(chat): honor the exact id and name filters when listing chats by @marmar9615-cloud in #19962
  • feat(vastbase): add Vastbase G100 document engine support with go by @FengYikai710 in #20419
  • Fix: remove raptor and graphrag when migrate parse config by @Lynn-Inf in #20599
  • feat(chat): add Search1API web search provider by @fatwang2 in #20561
  • fix(agentic): give web evidence stable content identities by @quifox in #20547
  • fix(agent): accept a string delimiter in StringTransform merge mode by @ekalinin in #20331
  • fix(web): regroup model instance card actions into the header by @dcc123456 in #20636
  • fix(web): keep API key modal within viewport with sticky table header by @dcc123456 in #20635
  • fix(canvas): send a status message without feeding it to the next node by @ondery in #20600
  • Refactor : Clean up compose and env file by @wangq8 in #20640
  • fix(agent): reject webhook body read failures by @T1mn in #20618
  • fix(agent): stop polling expired webhook traces by @T1mn in #20619
  • docs: mark macOS as temporarily unsupported for the Go backend by @xugangqiang in #20642
  • Doc: macOS(arm64) not supported now by @JinHai-CN in #20644
  • fix(ingestion): persist and honor global vision settings by @jay77721 in #20474
  • fix(api): persist tag_kwd on Go AddChunk (#20138) by @Harsh23Kashyap in #20146
  • feat(parser): unify BuiltIn and Pipeline parser selection into one dropdown by @xugangqiang in #20603
  • fix(agent): render and constrain generated PDF content by @6ba3i in #20604
  • chore(deps): bump music-metadata from 11.13.0 to 11.16.1 in /internal/channels/whatsapp-gateway by @dependabot[bot] in #20643
  • fix(infinity): restore deprecated knowledge-compilation columns in the chunk mapping by @qinling0210 in #20646
  • Fix: remove duplicate list function by @Lynn-Inf in #20648
  • feat(agentic_rag): add metadata_search tool to eino smart-reasoning path by @Caobaining1 in #20645
  • Bump Infinity to v0.7.4 by @qinling0210 in #20652
  • fix(web): temporarily hide the verify button on the draft model card by @dcc123456 in #20653
  • Fix: check retrieval node required fields by @Lynn-Inf in #20638
  • fix(web): remove the failover models field from chat settings by @yuzhichang in #20661
  • fix(agent): preserve attachment-only Message output by @6ba3i in #20655
  • fix(mcp): prevent managed token exposure by @6ba3i in #20639
  • fix(wiki): remove duplicate Markdown storage by @QAQadws in #20597
  • fix(name): unify case-insensitive entity name dedup and rename handling by @flyingwhit in #20594
  • fix(web): disable the create-key button once 16 API keys exist by @dcc123456 in #20665
  • Doc: fix readme by @JinHai-CN in #20663
  • fix(parser): honor the PDF page range for the VLM and plain_text methods by @devYRPauli in #20336
  • fix(chunk): persist tag_kwd on Go UpdateChunk (#20138) by @Xavierando in #20536
  • Docs: fix docker ps/logs by @wangq8 in #20671
  • Fix: Modify the style of the "Built-in" label in the ParserSelect dropdown menu. by @cike8899 in #20669
  • Remove opencloud9 support, since Go version is released by @JinHai-CN in #20674
  • feat(agent): add Search1API search and crawl tools by @fatwang2 in #20666
  • fix: implement CheckConnection and OCRFile for local MinerU driver by @why5602337 in #20459
  • fix(embedding): reject mismatched provider response counts by @kassdinzheng123 in #20662
  • fix(nlp): restore Infinity-safe escaping for synonyms and special chars by @li872 in #20572
  • Fix: The introductory information for the built-in parsing methods is no longer accessible. by @cike8899 in #20675
  • fix(parser): pass do_ocr and pdf_backend to Docling Serve by @BilalAtique in #20529
  • Fix: delete not found message should report error by @wangq8 in #20678
  • Fix: remove rank in mws config by @Lynn-Inf in #20676
  • perf(docker): optimize MySQL healthcheck parameters to eliminate startup delay by @MOHITPRADHAN35 in #20482
  • Update docker compose base for mysql by @JinHai-CN in #20682
  • fix: enforce dataset access for document detail and metadata summary by @QAQadws in #20649
  • fix(deepdoc): correct DLA channel order (BGR->RGB) and keep per-class NMS to match Python serving by @xugangqiang in #20637
  • fix(agentic_rag): number citations from a per-turn evidence registry by @Caobaining1 in #20677
  • Refactor: refine chat message to use question only by @wangq8 in #20654
  • fix: correct ingestion retry handling and enforce chunk permissions by @QAQadws in #20658
  • CI: optimize go build cache by @myf-bee in #20684
  • fix(parser): drop the BOM and decode JSON and Markdown input in Go parsers by @huiq777 in #20225
  • fix(agent): allow Invoke variables to use a reference or a literal by @pangwangshu in #20634
  • Doc: remove dead links to deleted Supported Models page by @shrxyas-246 in #20692
  • feat(parser): support genuine BIFF8 XLS workbooks by @Ricardo-M-L in #20573
  • refa: centralize Go resource permissions by @buua436 in #20657
  • CLI file timestamps reject offsets and discard milliseconds by @Taranum01 in #20491
  • fix(parser): resolve EPUB manifest hrefs as URLs by @huiq777 in #20228
  • feat(mcp): allow private MCP server hosts via deployment allowlist by @techpreneur-a in #20571
  • feat(web): show built-in pipelines in the agents list as read-only items by @xugangqiang in #20656
  • Fix: agent code exec test run does not work by @wangq8 in #20705
  • fix(dataset): reject an ingestion log_type other than dataset or file by @marmar9615-cloud in #19961
  • perf(knowledge_compiler): opt out of real-time retry backoff in failure-path unit tests by @xugangqiang in #20708
  • fix(deepdoc): stop PDF caption merge from dropping text and figures by @jay77721 in #20650
  • Fix: Limit the pipeline canvas to a maximum of one extractor operator. by @cike8899 in #20711
  • fix(parser): run image VLM when selected as parse_method by @flyingwhit in #20668
  • Go: refactor by @JinHai-CN in #20717
  • feat: expose dataset connector sync logs by @buua436 in #20704
  • Go: run first scheduled task immediately on Start instead of waiting one Interval by @JinHai-CN in #20714
  • CLI dataset document pagination is sent as HTTP headers by @Taranum01 in #20493
  • Fix: On the multi-model comparison page, the "Apply," "Delete," and "Add" buttons should be disabled whenever any of the models is generating output. by @cike8899 in #20723
  • Fix: ensure list datasetConnector result distinct. by @Lynn-Inf in #20710
  • refa: centralize agent canvas permissions by @buua436 in #20709
  • fix(wiki): tokenize GetWikiGraph keyword to match pre-tokenized content_ltks by @qinling0210 in #20722
  • fix(deepdoc): judge PDF header/footer and TOC removal against the full document instead of the parsed page subset by @flyingwhit in #20719
  • fix(chat): cap TTS text at 500 characters instead of 500 bytes by @marmar9615-cloud in #20344
  • fix(agent): cap Wikipedia extracts by runes instead of bytes by @marmar9615-cloud in #19959
  • fix(chunk): count merge budget in runes, not bytes by @kassdinzheng123 in #20690
  • fix(docker): use the SeekDB 1.4 data directory by @hnwyllmm in #20732
  • Go: refactor by @JinHai-CN in #20706
  • fix(agentic): preserve final answers after conversation reloads by @6ba3i in #20718
  • feat(go): MonkeyOCR env keys and tenant provisioning by @Souravrajvi0 in #20468
  • CI: optimize, Change the ci build global lock to a host lock by @myf-bee in #20729
  • Feat: Display the built-in pipeline canvas in read-only mode. by @cike8899 in #20742
  • fix(dataset): treat empty avatar as a clear signal on update by @flyingwhit in #20724
  • fix(agent): write version_title in CreateAgentSession so Explore sessions keep it by @flyingwhit in #20739
  • Fix: move model_provider function to separate file by @Lynn-Inf in #20743
  • fix(retrieval): search each tenant index once to stop duplicate chunks by @vedjaw in #20731
  • docs(engine): remove stale README by @QAQadws in #20398
  • refactor: Update entrypoint.sh by @JinHai-CN in #20744
  • Fix/syncer timer generation guard by @poijygfdyy in #20725
  • feat(parser): emit structured TableData instead of HTML for tables by @xugangqiang in #20741
  • fix(parser): drop hard image edge/pixel limit, add opt-in raster byte ceiling by @xugangqiang in #20745
  • feat(dataset-overview): show datasource sync logs on dataset logs page by @dcc123456 in #20738
  • docs: update migration by @Yannnnnnny in #20476
  • feat(setting-model): disable XunFei Spark provider configuration by @dcc123456 in #20753
  • Go: fix xunfei model by @JinHai-CN in #20757
  • Go: refactor model provider by @JinHai-CN in #20771
  • Go: use unified ToInt by @JinHai-CN in #20772
  • fix(pageindex): stop sampling the structure graph and emit rows in source order by @qinling0210 in #20750
  • Feat: Add a toggle switch for the auto-question feature of the extractor operator. by @cike8899 in #20748
  • fix(retrieval): preserve metadata filters during agentic thinking by @6ba3i in #20707
  • fix(agent-logs): persist failed execution status by @6ba3i in #20752
  • refactor(parser): split image OCR into an independent switch by @jay77721 in #20683
  • fix(chunker): preserve punctuation in general TXT and DOCX chunks by @jay77721 in #20751
  • fix(chat): preserve images across reasoning and conversation history by @6ba3i in #20728
  • docker compose: add enable admin parameter by @JinHai-CN in #20791
  • refactor by @JinHai-CN in #20783
  • docs: fix dead links to pages removed in the Go rewrite by @12britz in #20758
  • fix(web): restrict spreadsheet parser methods to supported options by @jay77721 in #20749
  • Fix Redis INFO memory reporting for Tair-compatible services by @Souravrajvi0 in #19905
  • Go: handle llama-server over-limit rejections in embed and rerank by @yuzhichang in #20784
  • Go: refactor by @JinHai-CN in #20795
  • Fix: The general parsing method for legacy data at the file level fails to display. by @cike8899 in #20794
  • docs: Removed the description of API_PROXY_SCHEME and synchronized the README files by @Yannnnnnny in #20756
  • doc: update development.md by @JinHai-CN in #20799
  • Go: unify tofloat by @JinHai-CN in #20797
  • Refactor commends and lib by @JinHai-CN in #20801
  • Fix: Validate the value of the 'topn' field on the dataset configuration page when the 'auto keyword' feature of the extractor operator is enabled. by @cike8899 in #20805
  • docs: fix stale and misleading links by @Yannnnnnny in #20804
  • fix: S3-compatible storage path-style addressing by @JinHai-CN in #20802
  • Fix: update by id without first by @Lynn-Inf in #20755
  • feat(ingestion): add progress fraction reporting to VisionEnhancement and cropImageChunks by @jay77721 in #20800
  • refactor: remove unused scripts by @JinHai-CN in #20813
  • chore(ci): remove obsolete sandbox plumbing by @6ba3i in #20811
  • refactor(sandbox): remove obsolete configuration by @6ba3i in #20809
  • Fix: Even though a vision model has been configured, uploading a video still triggers a notification stating that no vision model is configured. by @cike8899 in #20816
  • docs: sync the Elasticsearch-only vm.max_map_count step across README translations by @yuee3 in #20736
  • docs(sandbox): correct external deployment guidance by @6ba3i in #20810
  • Fix Agentic RAG does not do web search by @wangq8 in #20793
  • fix: cap chunk list page size when filtering by chunk ids by @dcc123456 in #20817
  • Fix: Agent - Start Component - When an input variable is set to optional: if a value is entered and then deleted, clicking "Next" triggers a "required field" error. by @cike8899 in #20820
  • fix(chunker): preserve PDF and Markdown sentence delimiters by @jay77721 in #20812
  • fix(parser): keep local MinerU images, positions, and API URL by @Souravrajvi0 in #20721
  • refactor(deepdoc): move testdata fetch into native and relocate script to tools/scripts by @xugangqiang in #20822
  • fix(ingestion): preserve typed metadata during merges by @jay77721 in #20808
  • fix(infinity): make pool max idle connections configurable via INFINITY_POOL_MAX_IDLE by @qinling0210 in #20831
  • fix(deepdoc): keep an attached figure caption on its own line by @jay77721 in #20819
  • docs: fix incorrect image references by @Yannnnnnny in #20835
  • Fix: agentic rag use the previous answer when LLM is bad by @wangq8 in #20829
  • fix(ingestion): preserve spreadsheet identity in rebuilt chunks by @jay77721 in #20825
  • Revert "fix(cli): drop PENDING as a parameter for message queue commands" by @flyingwhit in #20823
  • fix(chunker): slice General split positions by visible-text ratio by @jay77721 in #20827
  • fix(memory): keep unparseable invalid_at null so memories stay valid by @anneheartrecord in #20773
  • Refactor: No more reference files by @wangq8 in #20837
  • fix: unstable tests by @JinHai-CN in #20838
  • Fix: No connecting line appeared after adding an iteration operator following a conditional operator. by @cike8899 in #20833
  • Fix: handle int tenant_*_id by @Lynn-Inf in #20830
  • fix(config): mask credentials in Go startup configuration logs by @QAQadws in #20796
  • fix(web): recognize Lexical-normalized artifact links for wiki naviga… by @dcc123456 in #20828
  • fix(nlp): re-enable the datasetnav integration test in the integration CI job by @Caobaining1 in #20740
  • fix(deepdoc): clamp TSR/DLA boxes to image bounds at the analyzer exit by @xugangqiang in #20836
  • fix(web): make Google Drive "My Drive emails" optional for folder-only sync by @errmakov in #20528
  • feat(parser): support MinerU 4.0 V1 HTTP API by @Souravrajvi0 in #20712
  • fix: add buffer-length check in text.c (CWE-120) by @anupamme in #20803
  • fix(file): shorten long URL-upload filenames without dropping the extension by @Jinzhengxu in #20814
  • fix(es-to-oceanbase): preserve array values during JSON encoding by @lorenzozanee in #20695
  • Fix xunfei link by @JinHai-CN in #20849
  • fix: correct multi-tenant retrieval scope by @QAQadws in #20824
  • feat(parser): require a text source for images when vision enhancement is off by @jay77721 in #20839
  • fix(agent): finish responses within configured tool budgets by @6ba3i in #20847
  • fix(file): reject moving a folder together with its descendants by @flyingwhit in #20848
  • fix(infinity): drop row keys that are not table columns by @devYRPauli in #20513
  • Add Demon Route as an LLM provider by @demonroute2 in #20853
  • fix(deepdoc): preserve PDF captions labeled as references by @lorenzozanee in #20767
  • docs: Update the CLI documentation to reflect the actual implementation. by @Yannnnnnny in #20821
  • fix(service): keep citations and reference for a no-answer reply by @Caobaining1 in #20862
  • fix(knowledge_compile): preserve structure merge metadata by @lorenzozanee in #20858
  • fix: accept .markdown files in upload whitelist by @heranliu in #20850
  • fix(web): allow pipeline methods to save in change-parser dialog by @dcc123456 in #20864
  • Fix: Remove the switch for deleting the original doc/html/markdown directories in the parser operator. by @cike8899 in #20866
  • Feat/migrates navigation tools form old mode to agentic mode by @Caobaining1 in #20735
  • fix(sandbox): reject nonpositive local output limits by @6ba3i in #20874
  • refactor(parser): bake merged-cell geometry into the structured TableData grid by @xugangqiang in #20832
  • fix(api): web API update by @JinHai-CN in #20870
  • docs: standardize parameter names in http api docs by @Yannnnnnny in #20868
  • Refactor: move rag prompts by @wangq8 in #20873
  • feat: migrate resource checks to permission layer by @buua436 in #20792
  • Feat: Access the "Naive" and "Agentic" menu options on the chat page. by @cike8899 in #20877
  • fix(sync): preserve published blobs and retry incomplete revisions by @dakjdakd in #20627
  • fix(artifact): reject non-integer top_n in wiki graph route (#20006) by @RerankerGuo in #20483
  • fix(sync): reject incomplete Google Drive snapshots before pruning by @dakjdakd in #20629
  • fix(syncer): honor ALLOW_ANY_HOST for private database hosts by @ondery in #20574
  • fix letter case issue by @JinHai-CN in #20878
  • update go.mod by @JinHai-CN in #20879
  • fix(admin): restore saved settings when switching sandbox executors by @6ba3i in #20863
  • fix: seed dataset owner membership in artifact tests by @buua436 in #20886
  • fix(syncer): fail closed on non-positive connector refresh freq by @jay77721 in #20871
  • Fix: Add SANDBOX_EXECUTOR_MANAGER_ALLOW_UNAUTHENTICATED by @wangq8 in #20889
  • fix(ingestion): fetch a PDF source before image admission in VLM enhancement by @jay77721 in #20846
  • fix(syncer): use the S3-compatible client for S3 sources set to compatible mode by @marmar9615-cloud in #20342
  • fix(sandbox): compile the local provider on macOS and the BSDs by @wenqiw777 in #20334
  • Go model: fix MinerU local server URL decoding by @JinHai-CN in #20891
  • fix(agent): fall back to the default email when PubMed email is blank by @ekalinin in #20509
  • fix(agent): honor sampler enable flags in Go LLM and Agent components by @ekalinin in #20259
  • feat(web): preview source document from chunk reference images by @dcc123456 in #20895
  • fix(web): expand citation source-locate to all previewable file types by @dcc123456 in #20898
  • Fix: Connecting other nodes after the message node is prohibited. by @cike8899 in #20897
  • fix: allow declining pending tenant invitations by @buua436 in #20896
  • fix(chunker): preserve separators between JSON items by @lorenzozanee in #20857
  • fix(chunker): preserve whole lines in PDF text previews by @jay77721 in #20890
  • fix(storage): prevent MinIO client reconnect races by @6ba3i in #20888
  • fix(agent): apply cross_languages in Go Retrieval memory mode by @errmakov in #20328
  • Refactor: log name without port by @wangq8 in #20905
  • Refactor: Delete not used APIs by @wangq8 in #20894
  • fix(web): drop UNSTART running status and align status labels by @dcc123456 in #20904
  • Fix: The document-level graph search box is not displaying correctly. by @cike8899 in #20899
  • Fix: apply docID filter by @Lynn-Inf in #20908
  • fix: scope ingestion documents and reused tasks to the dataset by @QAQadws in #20806
  • Fix: Update the prompt text for file types not supported by the current pipeline. by @cike8899 in #20911
  • fix(harness): return empty iterator when seek clock exceeds event log by @poijygfdyy in #20726
  • docs(chat): describe the two remaining thinking modes by @Caobaining1 in #20916
  • Fix: Validate the parameters of the variable assignment operator. by @cike8899 in #20915
  • fix(chunk): keep the chunk list inside the requested document scope by @qinling0210 in #20914
  • refactor(deps): split dependency download into two Python scripts by @xugangqiang in #20861
  • refactor(cmd): parse CLI flags with the stdlib flag package by @xugangqiang in #20876
  • fix(ingestion): report knowledge compiler progress fractions by @jay77721 in #20902
  • refactor: remove the low/medium/high/ultra thinking modes by @Caobaining1 in #20910
  • fix(web): support audio and video citations in source locate by @dcc123456 in #20932
  • fix(agent): process all configured Message inputs by @6ba3i in #20936
  • Fix: Changing the variable assignment operator results in the error: c.trim() is not a function by @cike8899 in #20937
  • CI : Optimize go cache, optimize runner space growing too quickly. by @myf-bee in #20931
  • fix(agent): reject marker-only citation grounding responses by @li872 in #20685
  • fix(agent): return the StringTransform split result as a generic list by @ekalinin in #20920
  • fix(web): only close rename modal after a successful save by @dcc123456 in #20940
  • fix(web): take mind map trigger out of absolute positioning by @dcc123456 in #20939
  • fix(dataset): preserve pipeline child chunk delimiters by @6ba3i in #20942
  • fix(web): auto-scroll to newly linked data source by @dcc123456 in #20935
  • fix(web): filter empty metadata values in dataset meta_fields count by @jay77721 in #20933
  • fix(kvrocks): surface transport errors from SetNX and DistributedLock.Acquire (#20660) by @Harsh23Kashyap in #20774
  • fix: disable max token defaults in agent templates by @buua436 in #20946
  • Feature: add knowledge_compilation_scope by @wangq8 in #20917
  • fix(searchbot): honor the search app's saved retrieval config by @qinling0210 in #20944
  • Fix: prevent database query when no-data by @Lynn-Inf in #20938
  • Fix: handle redundant v1 in base_url by @Lynn-Inf in #20952
  • Fix LLM setting for 3 search config by @wangq8 in #20955
  • fix(infinity): make memory messages work on the Infinity engine by @qinling0210 in #20954
  • Fix: Validate the SofyaSearch API key in the agent tool. by @cike8899 in #20945
  • Fix: remove deprecated code by @Lynn-Inf in #20970
  • fix(harness): preserve message fields and isolate agentic content by @QAQadws in #20869
  • fix: accept ReAct completion on the final iteration by @QAQadws in #20875
  • fix(agent): propagate provider failures before downstream execution by @6ba3i in #20968
  • Render table chunks as HTML on Text at the chunker by @xugangqiang in #20918
  • refactor(config): resolve CLI/env/file precedence in one place with final-value validation by @xugangqiang in #20934
  • Add Slovak & Czech language dataset support by @martincivan in #19013
  • fix(agent): cap generated document names at 180 characters and 255 bytes by @Jinzhengxu in #20909
  • fix: clarify extractor timeout errors by @buua436 in #20983
  • fix(web): sync excel source-locate scrollbar and fix highlight contrast by @dcc123456 in #20960
  • Fix: Remove chunk metadata from the chat and search pages. by @cike8899 in #20985
  • fix(chunker): stop charging the replicated table header against the split budget by @jay77721 in #20845
  • test(api): migrate REST API coverage to Go by @6ba3i in #20991
  • fix(aliyun): transcribe audio via DashScope multimodal-generation by @12britz in #20975
  • fix(chunker): keep ancestor headings on every cap-split hierarchy piece by @jay77721 in #20953
  • fix: scope chunk deletion to route dataset and add Go API coverage by @buua436 in #21000
  • docs:Improve local model deployment and provider configuration docs by @Yannnnnnny in #21002
  • Align code with EE by @JinHai-CN in #21007
  • fix: Removed broken supported_models links from all UI translations by @Yannnnnnny in #21009
  • fix(agent): make single-component debug match real-run state by @flyingwhit in #20807
  • fix(chunk): list parent chunks after source chunks on every doc engine by @qinling0210 in #20990
  • fix: clarify compiler timeout errors during ingestion by @buua436 in #21011
  • fix: preserve named Begin inputs in agent bot runs by @buua436 in #21014
  • fix(agent): preserve empty attachment brackets by @6ba3i in #21016
  • Refactor: split ingestion.go by @Lynn-Inf in #21017
  • fix(web): resolve cited chunk by marker key in reference image carousel by @dcc123456 in #21012
  • fix(web): remove a provider-instance draft right after it saves by @dcc123456 in #20947
  • fix(chunk): honor pagination when filtering by entity chunks by @QAQadws in #21015
  • Fix: Separate the hierarchy and group configurations for the title chunker into distinct fields. by @cike8899 in #21013
  • Fix: Use TEI_BASE_URL then TEI_HOST by @wangq8 in #21021
  • fix(dao): migrate Python-era flat ingestion config onto pipeline components by @jay77721 in #20986
  • docs: add tag migration description by @Yannnnnnny in #21020
  • Fix: Prevent the left-side handle of a node on the canvas from serving as a starting point for connections to other nodes. by @cike8899 in #21019
  • fix(parser): keep short-line body pages out of PDF TOC removal by @jay77721 in #20988
  • fix(parser): forward caller ctx through PDF TCADP path by @Taranum01 in #17673
  • Remove dead ParseResult.HTML field and redundant EmailParser json-path Text by @xugangqiang in #20992
  • fix(parser): inherit merged data cells, strip invisible chars, trim dataless columns by @jay77721 in #20996
  • test(api): migrate memory and message endpoint coverage by @6ba3i in #20995
  • test(api): migrate OpenAI compatibility and plugin tool tests by @6ba3i in #20999
  • test(api): migrate and improve retrieval API coverage by @6ba3i in #21004
  • fix(agent): remove empty tool from stock research template by @6ba3i in #20987
  • fix(chunk): expose per-chunk available in the chunk list by @qinling0210 in #21022
  • Fix: A user-friendly presentation of the indexer's search method by @cike8899 in #21024
  • refactor(cmd): extract server runners into internal/server/* packages by @xugangqiang in #20883
  • fix(search): delete the Search App row like Python does, drop unused soft-delete helpers by @qinling0210 in #21005
  • refactor(parser): type ParseResult.JSON as []DocItem boundary adapter by @xugangqiang in #20998
  • Fix(dataset):auto parse unsupported upload by @flyingwhit in #21018
  • fix(ingestion): scope extractor timeout to single LLM call by @jay77721 in #21003
  • Fix:tokenizer pool race condition by @flyingwhit in #20903
  • fix: cancel GCS upload context when Put returns by @QAQadws in #20885
  • Refactor: move cache to transaction by @Lynn-Inf in #21023
  • fix(files): handle vanished folders on list requests instead of returning 500 by @flyingwhit in #20892
  • fix(examples): use PATCH for chunk update in chunk_example.sh by @jrlprost in #21037
  • fix(infinity): keep one row per chunk id on insert by @qinling0210 in #21043
  • Dev: keep EE license by @wangq8 in #21048
  • fix(parser): grow truncated rows to host inherited merge values by @jay77721 in #21044
  • Fix CLI output truncation by @JinHai-CN in #21052
  • Fix: Add a tooltip to the 'tag' field of the extractor operator. by @cike8899 in #21047
  • fix: update tenant model max tokens migration version by @buua436 in #21045
  • Fix: rm dead code in kb.go by @Lynn-Inf in #21039
  • Fix: The greater-than sign in the message cannot be displayed. by @cike8899 in #21055
  • Build: add multi-platform native build support (linux/arm64, darwin) by @xugangqiang in #19247
  • docs: unify sandbox docs by @Yannnnnnny in #20951
  • Feat/agentic kg outline and caller identity by @Caobaining1 in #21053
  • Go: update EE interface by @JinHai-CN in #21056
  • test(api): migrate Python dataset API coverage to Go by @6ba3i in #21063
  • test(api): migrate and improve task route coverage by @6ba3i in #21060
  • test(api): migrate and improve session API tests by @6ba3i in #21057
  • test(api): migrate and strengthen system API tests by @6ba3i in #21059
  • Fix: set tag to v1.0.0-rc2 by @Lynn-Inf in #21061
  • dockerfile and release add arm64 build by @myf-bee in #21067
  • Go: remove unused APIs by @JinHai-CN in #21073
  • test(api): migrate Python document API coverage to Go by @6ba3i in #21066
  • feat(infinity): implement SearchByRegexp for the agent grep tool by @qinling0210 in #21068
  • Docs: Update version references to v1.0.0-rc2 in READMEs and docs by @asiroliu in #21069
  • Docs: 1.0.0-rc2 release notes by @wangq8 in #21064

New Contributors

Full Changelog: v1.0.0-rc1...v1.0.0-rc2

Don't miss a new ragflow release

NewReleases is sending notifications on new releases.